Novel batch active learning approach and its application to synthetic aperture radar datasets

Tue, 13 Jun 2023 00:00:00 +0000

The paper introduces a two-stage strategy for making batch active learning nearly as accurate as sequential querying while substantially reducing the number of classifier updates. Dijkstra’s Annulus Core-Set (DAC) first constructs a representative candidate set; LocalMax then selects a diverse batch by enforcing local maxima of the acquisition function on the data graph.

The resulting pipeline combines transfer-learned image features, graph-based semi-supervised learning, DAC, and LocalMax for synthetic-aperture-radar target classification. On FUSAR-Ship and OpenSARShip, it retains the accuracy of sequential active learning with speedups that scale with batch size and outperforms the evaluated CNN-based baselines.

Dijkstra's Algorithm | Bohan Chen's Personal Webpage

Novel batch active learning approach and its application to synthetic aperture radar datasets